#108 - N-Gram Analysis for Amazon PPC: How to Reduce ACOS with Negative Phrase Match Keywords

#108 - N-Gram Analysis for Amazon PPC: How to Reduce ACOS with Negative Phrase Match Keywords Speaker 1: You may have heard about N-Grams for Amazon PPC. It's an advanced method of search term analysis that allows you to uncover hidden and wasted ad spend so that you can lower your ACOS. Speaker 2: If you haven't heard about N-Grams, no problem. Today, we're breaking down the definition and showing you how to make full use of this advanced tactic to improve your Amazon PPC. Speaker 1: Alexa, play That Amazon Ads Podcast. Unknown Speaker: Which one would you like to hear? Speaker 1: The best one. Unknown Speaker: Okay, now playing That Amazon Ads Podcast. These gentlemen are completely changing the game. Speaker 2: After listening to That Amazon Ads Podcast, my ads are finally profitable. Unknown Speaker: I also heard they're pretty cute. Speaker 1: All right guys you may know that we have that Amazon ads master class which is a online course the engine I created last year it's basically the podcast information but. Distilled and on steroids. So it's streamlined, faster, more intense content. You can check it out. There's a link in the description. We are overdue. We are supposed to upload a few masterclass videos to YouTube, just some of the teasers. So be on the lookout for those. We're going to start uploading just a few of the lessons for free. Make sure you subscribe so you don't miss it. One of those lessons is the N-Grams tutorial, which I don't know if that's coming out Before this episode or after this episode. So look for that. But we're gonna do a real quick screen share and just show you very quickly what is going on in the Google Sheet that Andrew created. So Andrew, we've got this sheet. Why don't you just give us like the 10, 15 second tutorial for how to use this Google Sheet and then we're gonna show you guys a better, faster way. Speaker 2: Yeah, so basically to use this sheet, what you need to do is go download a search term report, upload it to this sheet, and then it will parse everything out for you and help you start to analyze these different sequences of words. It's pretty straightforward. In the lesson, we'll walk you through exactly how to do all that. Speaker 1: Yeah. So just take that search and report, last 30 days, 60 days, grab the impressions, click spend orders, add sales, just paste it in the same form that you see here. And then the Google Sheet will do the magic and break everything down. Now, before we actually Walk through that sheet. I totally forgot to offer the definition of N-Grams. So, Andrew, do you want to tell people what the definition, what the classic legal definition of an N-Gram is? Speaker 2: Yeah, this is from ChatGPT, so it's official. An N-Gram is a contiguous sequence of N words for a given phrase or text. And I was confused, like, what's that N mean? That just means number of words. So we're basically looking at a phrase or individual words and Yeah, that's what you have in your search term report. Anyway, anything else to add to that definition? Speaker 1: I think contiguous might need a definition. It means sharing a common border or touching. So the example that we have is if you have the sentence, Okay, I'm saying sentence, but imagine these are search terms, right? You have the sentence, the search term, the cat runs fast. A unigram or monogram, a one-gram is, when we say N-gram, the N is just a number. So you can have a one-gram, a two-gram, a three-gram. That's what the N is signifying. So in a monogram or a unigram with one letter, you're breaking that search term out into the, as one word, cat, another word, runs, another word, fast. And then bigram or two would be you'd have the cat and cat runs and runs fast. Because you're breaking all of that out. And then the trigram or three would be the cat runs and then on the first half and then cat runs fast on the third. And you can see that with that contiguous definition, all of those words are directly adjacent and connecting to each other. But when it comes to Amazon PPC, those search terms, so it's basically functioning like phrases, right? So you've got the phrase, the cat runs versus the phrase cat runs fast. But when we do our N-Gram analysis, we prefer to use a little bit more of a of a modified broad match definition, which allows the reordering of words. So you can have fast cat runs, for example, would still be included in a three gram. So it's three words. That are within that that sentence or search term, but not necessarily they don't have to be directly adjacent and in that exact same order. Andrew, can you think of any like I'm trying to I couldn't think off the top of my head, but any. Let's talk a little bit about what you're doing. What are some of the examples that come to mind of where allowing that rearrangement can actually be kind of helpful? Speaker 2: Yeah. So I have a client who sells goggles and we did this analysis and we noticed that prescription kept showing up in a lot of search terms, but it wasn't always like adjacent. It wasn't like prescription goggles always. It was like goggles for swimming, prescription, things like that. So having that flexibility to reorganize things, was something that they did not sell that wasn't performing super well. So having that reordering definitely helped. Speaker 1: That makes total sense. And this is where this starts becoming very valuable, right? When people do their search term analysis, they normally are only looking at search terms. So you can have tons of really long tail search terms. Like, and when we say long tail, we're saying like the word count in that term is like four or five, six words long. And you've got a thousand different variations of these long tail search terms that individually, they all only have like 50 cents of spend, a dollar of spend, nothing big. But in that same list of a thousand search terms, you have the word prescription and goggles appearing over and over and over again in any kind of order or rearrangement. So you can have prescription goggles for swimming or goggles for swimming with prescription or goggles prescription, whatever. So that's where the bigram is going to break apart all of the search terms in your entire search report and break out those individual words and couple them back together to aggregate that data. So on this Google Sheet, you can see the monogram version. Is taking every individual word and breaking it out and showing you exactly what the performance metrics are like versus the bigram, which is now taking two of these. So, you know, you can imagine that like, you know, peanut butter, for example, you're probably not gonna get people typing in like butter peanut, you know, things like that. But they could be rearranged in a few ways. And then trigram just gets you one, layer deeper. The purpose of these is to uncover negative phrase match opportunities where you just can't normally find this data aggregated at this level when doing, yeah, when just looking at a simple search term report. Anything to add here, Andrew? Speaker 2: Yeah, a lot of times you won't see it on the individual search term level in some of these things like prescription goggles, like maybe that only shows up once or twice here and there, but across the entire account, whenever you take all that data aggregated across those individual words and individual phrases, it starts to add up and shows you a little bit of a trend whenever you've rolled it all up. So you may not catch it if you're just going through looking for negative opportunities within your search term reports, like especially if it has to have reached a certain level of spend and things like that. Sometimes these things are really low spend, but overall on aggregate, whenever you start to combine the performance across the entire account or group of campaigns, it really starts to become offensive and dictate some of your performance. So you're able to see it a little bit more clearly whenever you're able to aggregate it up across all those little instances where that word shows up. Speaker 1: A hundred percent. Now, let's switch over to our preferred method because the You know, the search term report, this Google sheet works great. Only downfall is your search reports can only get like 60 days of data. And you're also, it can be a pain to have to download it, upload it here, and then go find those campaigns. Speaker 2: And then like, if we could just sort this by like- Sometimes Google sheets will crash too. Like if you have too many search terms. Speaker 1: So like, this is a good example where it's like you see fermented, you're probably not getting the search term fermented. It's not appearing as like a standalone search term, but it is appearing in longer tail keywords and is getting averaging 350% ACOS. So you might just want to like, you can, the next thing you have to do is go back to your search report, filter out for contains fermented to see where that's appearing and try to, you know, find those campaigns and then add the negatives. We're going to be using AdLabs because what we can do here is I've hidden the actual search term columns just to maintain some confidentiality here. But you can do a whole like year to date or two years data, whatever timeframe you want. And you can select all these search terms. You see that we've got 42,000 search terms here. And then just click this little N-Grams icon, which is going to open up this window. And now you can enter the number of words or ends, N-Grams you want to count to. So you could do a monogram versus a bigram versus trigram. So let's just start with a monogram. It's going to be probably a lot. So it calculates it. That's 4,000 unique words. It's actually kind of interesting. So out of the 40,000 unique search terms, there's 4,600 unique We're going to talk about individual words and we can now adjust some filters here. So maybe we want to do, I already got this sales equals zero. So just look at the stuff where sales equals zero and let's sort by the highest spend. I think we already sorted like that. So now we can start to analyze. This is kind of weird. That's a very strange, long tail search term. I'm surprised it got so much spend. I have no idea how that I'm going to have to look into that one later. Money. All right. Not doing well. Hanging. Not doing well. CCW, hang, plier. So you can start to California, like as far as I'm aware of this, this, brand doesn't have anything that's California specific. So it's appearing in five different search terms. If I hit this dropdown, I can see California framing hammer. So I can drop down and see what more specifically, which search terms and campaigns this is appearing in. Speaker 2: I'll just say this too. This works. We're here to talk about how you can get better whenever you have longer date ranges. Things start to become a little bit more apparent whenever you're using those longer date ranges. Stephen selected year-to-date data. You may not find these more offensive keywords and things spending a whole lot if you're looking at like a 30-day, even a 60-day timeframe. So being able to extend that date range back a lot further will really reveal a lot of these underlying issues here with your search term. So just wanted to note that. Speaker 1: Whenever there's a search term that you don't know what it means, like you don't know what people are looking for because I am not good with hammers, just go search it real quick, right? And it looks like now there's two actions I could take here. I could do negative exact California framing hammer, or I could go negative phrase on California framing hammer, or I can just negative phrase California which I mean in this case that's kind of where the end grams are supposed to help. So, just to be just to kind of, if you do it, it's a weird situation where all the search terms were identical. But when you start getting all these kind of weird things like I had one brand in the ceiling fan category. And when I did N-Grams, I noticed I was getting like thousands of search terms that would contain the word battery, where none of them were individually like bad performing search terms. But in aggregate, we were spending like $3,000 at like 800% ACOS on people that were searching for battery powered ceiling fans or battery replacement ceiling fans, all those kind of things. We're the same things we have are not battery powered. They're just you plug them into the grid. So that was a good opportunity to just negative phrase battery from 100% of our campaigns. And so in this case too, same with this, if we just, this has a negative phrase at the campaign level, ad group level, whatever, it's going to automatically add it to the campaigns that it appeared in. But I can also just expand that to all of my campaigns. If I just really want to make sure that We just are preemptively covering it from all other campaigns. So now it's been added to everyone. So we shall never again see California in our search terms, at least not on these campaigns. Obviously, if I launch new campaigns, I have to like, you know, add that negative. But starting from August into the rest of the lifetime of these campaigns, we're never going to see California appear. So that is how you can start trimming this down. I mean, in total, all this stuff is like, how much wasted ad spend is this? $7,500 of ad spend. It's down here where I'm getting that from. Isn't that a nice feature? So yeah, I mean, you start to get the idea of where you can optimize. The other thing too is you also want to look at high ACOS, right? So if ACOS is greater than, let's just start with like 100% or something like that. And we can sort this by highest day costs and start to do additional analysis. Mole. I don't know what this is. Let's check it out. Speaker 2: This is so good for just finding those irrelevant terms too. Like you're breaking it down individual terms. It's very easy to go through and see the stuff that's relevant to your product or not. We've talked about prescription goggles. You mentioned California. Now we're looking at mole, like just going through scanning through these and seeing, Does this even relate to the product? You mentioned batteries on your fans. Those are the obvious things that you're looking for. In general, the negative criteria that we're looking at is, is it relevant to the product or not? A lot of times you can tell that from the data. If it's not converting, it's got a lot of spend, may not be relevant. These are just a couple of filters you can use to help narrow that down and find those things that are irrelevant that could be ruining some of the performance within your account. Just to backtrack to your hammer example, a California framing hammer, it's just the shape of a hammer in case anybody was wondering. Speaker 1: This mole up here is apparently a competitor brand term. These aren't specifically competitor conquesting campaigns. I can probably come back here later and add that as a negative, but you see the point here. There's a ton of different stuff That I need to go through because like wallet, for example, we have one product that's kind of like a money clip holder, but it's like a very bad product that we're discontinuing. Obviously, it's not easy. You can kind of get a sense of the other products here are based on like more construction type stuff. So wallets are not our main thing. I'm going to go through here and just double check that that search term is not appearing in like my other non-wallet campaign. So there's just a lot of analysis that you can dig through here and try to find out what's going on. And now I can also switch this to Check for two grams, five grams and see how things change. So now we've got pouch bag, badger tool, EDC, blah, blah, blah, tools and whatever. So lots of ways to just manipulate, uncover, discover and try to find opportunities where things are slipping through the cracks that your normal search term filters can't find. This is basically just an advanced filtering system That is identifying deeper trends within your search reports for the purpose of negative phrase match keywords. Speaker 2: There are so many valuable pieces and insights that you can gain from doing this level of analysis using N-Grams. You're able to uncover those inefficiencies, spot opportunities for adding those negative phrase keywords, finding those root keywords that are really offensive within your account and start to clean some of that stuff up. If there was anything that you had questions on or that we didn't cover in this episode that you'd like us to cover, make sure you leave a comment. Make sure you like and subscribe this video and stay tuned for more episodes as we drop. Again, as Stephen mentioned, we've got the Masterclass exclusive videos coming out soon. So those will be really value-packed and value-driven. So stay on the lookout for those and we'll see you next time on That Amazon Ads Podcast. Speaker 1: Peace out. Unknown Speaker: I'm Andrew.